Extremal Domain Translation with Neural Optimal Transport

Abstract

In many unpaired image domain translation problems, e.g., style transfer or super-resolution, it is important to keep the translated image similar to its respective input image. We propose the extremal transport (ET) which is a mathematical formalization of the theoretically best possible unpaired translation between a pair of domains w.r.t. the given similarity function. Inspired by the recent advances in neural optimal transport (OT), we propose a scalable algorithm to approximate ET maps as a limit of partial OT maps. We test our algorithm on toy examples and on the unpaired image-to-image translation task. The code is publicly available at https://github.com/milenagazdieva/ExtremalNeuralOptimalTransport

PDF NeurIPS OpenReview Semantic Scholar

Cite

Text

Gazdieva et al. "Extremal Domain Translation with Neural Optimal Transport." Neural Information Processing Systems, 2023.

Markdown

[Gazdieva et al. "Extremal Domain Translation with Neural Optimal Transport." Neural Information Processing Systems, 2023.](https://mlanthology.org/neurips/2023/gazdieva2023neurips-extremal/)

BibTeX

@inproceedings{gazdieva2023neurips-extremal,
  title     = {{Extremal Domain Translation with Neural Optimal Transport}},
  author    = {Gazdieva, Milena and Korotin, Alexander and Selikhanovych, Daniil and Burnaev, Evgeny},
  booktitle = {Neural Information Processing Systems},
  year      = {2023},
  url       = {https://mlanthology.org/neurips/2023/gazdieva2023neurips-extremal/}
}